Sotaro Iijima

早稲田大学先進理工3年
AIアニメの会社AiToGet代表
再生医療とか老化とか好き
最近ブルーアーカイブを始めた

[deepbeepmeep @deepbeepmeep]

HunyuanVideo GP v4 is out, check it out !: fast mode to generate 4s of a 1280x720 video with multiple loras in only 4 minutes on a RTX4090, on the fly change of models, attention modes without restarting the app, ..
https://t.co/TXeapIgweG

[とりにく @tori29umai]

animagin4.0かfluxでanytestの後継的なもの、月須和さんが作ってくれたらうれしいんだけど、せっかく作り��公開して下さっているんだし、人任せじゃなくて自分でもやるかーーー（ぐてり）ってなっている
ようはこれを大量に作れば良いわけでしょ？・・・良く思いついて実行に至ったなぁ

[toyxyz @toyxyz3]

https://t.co/8neE6jraB1

[Vaibhav (VB) Srivastav @reach_vb]

NEW: Open Source Text/ Image to video model is out - MIT licensed - Rivals Gen-3, Pika & Kling 🔥

> Pyramid Flow: Training-efficient Autoregressive Video Generation method
> Utilizes Flow Matching
> Trains on open-source datasets
> Generates high-quality 10-second videos
> Video resolution: 768p
> Frame rate: 24 FPS
> Supports image-to-video generation

> Model checkpoints available on the hub 🤗

[camenduru @camenduru]

🖼 LVCD: Reference-based Lineart Video Colorization with Diffusion Models 🖌 Code is available now! 🔥🔥🔥

Thanks to Zhitong Huang ❤ Mohan Zhang ❤ Jing Liao ❤

🌐page: https://t.co/ANEvpN7jIq
🧬code: https://t.co/XOlm58MLDs
📄paper: https://t.co/KDA62R7EDk
🍊jupyter: https://t.co/hRGnrwi5Ua

[apolinario (poli) @multimodalart]

Editing facial expressions in real time now on @huggingface Spaces 👨‍🎤🔀

A Grog converted Cog image to Gradio running a ComfyUI backend - magic of open source 🤝

▶️ https://t.co/dxICPo9sYO

[動詞 @IMG_5955]

Flux.1の推論を高速化するComfyUIカスタムノードを作ってみました。ノードにモデルを通すだけで生成が35〜37%高速化できます！

DiT内の各Blockを指定して計算をスキップするっていう黒魔術も搭載したので、最適化の可能性は無限大です！！！
使ってみてね！！
https://t.co/k24mjXvzjx

[ComfyUI @ComfyUI]

https://t.co/eleSiei7Yh

[Sotaro Iijima @iss372]

ラフと線画データセット、ラフ→線画じゃなくて線画→ラフで学習すれば3Dモデルを使ったラフ→線画素材が無限に作れるようになるのか？

[AK @_akhaliq]

InstantDrag

Improving Interactivity in Drag-based Image Editing

discuss: https://t.co/dTHqxUTdSg

Drag-based image editing has recently gained popularity for its interactivity and precision. However, despite the ability of text-to-image models to generate samples within a second, drag editing still lags behind due to the challenge of accurately reflecting user interaction while maintaining image content. Some existing approaches rely on computationally intensive per-image optimization or intricate guidance-based methods, requiring additional inputs such as masks for movable regions and text prompts, thereby compromising the interactivity of the editing process. We introduce InstantDrag, an optimization-free pipeline that enhances interactivity and speed, requiring only an image and a drag instruction as input. InstantDrag consists of two carefully designed networks: a drag-conditioned optical flow generator (FlowGen) and an optical flow-conditioned diffusion model (FlowDiffusion). InstantDrag learns motion dynamics for drag-based image editing in real-world video datasets by decomposing the task into motion generation and motion-conditioned image generation. We demonstrate InstantDrag's capability to perform fast, photo-realistic edits without masks or text prompts through experiments on facial video datasets and general scenes. These results highlight the efficiency of our approach in handling drag-based image editing, making it a promising solution for interactive, real-time applications.

[Zuntan @Zuntan03]

簡単 Flux な EasyForge の flux_tool で、Civitai 週間一位の iNiverse Mix を高速化例

Geforce RTX 3060 12GB で高速 nf4 で一枚 30秒、高品質 Q8_0 で 45秒
https://t.co/VXNu6mX7Ea

画像は公式例、4枚目は Dev 用アニメ LoRA 併用。
NSFW も可
https://t.co/HgEGbxtDTP
https://t.co/rWhI4GuX46

[Zuntan @Zuntan03]

簡単 Flux な EasyForge に『Flux Dev 簡単クッキング』を追加。
LoRA 互換を保ちつつ Flux Dev 高速版を「自分好み」にマージするチュートリアルです。

成人向け作例
https://t.co/PMRYAoIzbT

解説
https://t.co/S3jgpDBvC2

アニメ NSFW モデル例
https://t.co/rAmGlyRcEY
https://t.co/gjh8kTIpkP

[toyxyz @toyxyz3]

https://t.co/UwUJPH7E0O

[toyxyz @toyxyz3]

https://t.co/WCb3PD5CPW

[toyxyz @toyxyz3]

https://t.co/KvrhUq0yUK

[Ostris @ostrisai]

Did a lot of testing on my LoRA training script for @bfl_ml FLUX.1 dev model. Amazing model! I think it is finally ready. Running smooth on a single 4090. Posting a guide tomorrow. Special thanks to @araminta_k for helping me test on her amazing original character and artwork.

[FrankChieng @kurtqian]

after some Flux Lora hyperparameter fine-tuning and 3000 steps,just load the lora model in ComfyUI,the human being aethetic image is much more better than SDXL,validation eval and generation portrait not bad, the LR set with 4e-4 and optimizer adafactor,it's the tricky part

[camenduru @camenduru]

🤯🦋 CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer 🔥 @runpod template

Thanks to CogVideoX Team ❤

📄paper: https://t.co/4zo5vWP3Wb
🧬code: https://t.co/D6DM0wyIcu
🍇runpod: please try it ��� https://t.co/r0L9UrZF5L

[花田十輝 @oitan125]

ヒットしたアニメオリジナル作品の１話もほとんどはこの構成。ここに書かれたことが明かになると、視聴者は作品を「理解した」となります。自分の最近の作品でいうと、サクナヒメの１話は綺麗にこの構成で、項目2の世界設定がやや特殊なので、ゆいに海を覗かせて台詞で補完。原作だともう少し後になる5もなるべく早い方がいいので、アニメでシーンを追加して１話に先に提示という感じ。興味ありましたら。

[toyxyz @toyxyz3]

https://t.co/77g5GkZ4OR